A statistical analysis of RNA folding algorithms through thermodynamic parameter perturbation

Area of Science:

• Computational biology
• Biophysics
• Molecular biology

Background:

• Computational RNA secondary structure prediction is a well-established field.
• Prediction algorithms rely heavily on experimentally determined parameters.

Purpose of the Study:

• To investigate the sensitivity of RNA structure prediction algorithms to variations in experimental parameters.
• To assess the impact of parameter uncertainty on prediction accuracy and ground state stability.

Main Methods:

• Analysis of parameter perturbation effects on structure prediction.
• Calculation of base-pairing probabilities within a thermal ensemble.
• Development and evaluation of a new stability measure based on parameter perturbation.

Main Results:

• A 30% false prediction rate was observed with parameter changes within experimental error margins.
• The ground state structure remained stable under parameter perturbation in only 5% of cases.
• Base-pairing probabilities are a useful, though imperfect, indicator of prediction reliability.

Conclusions:

• RNA structure prediction algorithms exhibit significant sensitivity to parameter variations.
• A novel stability measure using parameter perturbation offers insights but has limitations.
• Further research is needed to improve the robustness of RNA structure prediction models.